Llama and more
Meta
from $0.0175 / 1M
23 models · typical $0.10 · top 67.0% GPQA
Developer head to head · prices updated Oct 31, 2026
Meta vs OpenAI
Llama vs GPT: every model's API price, compared
Meta starts cheaper ($0.0175 vs $0.055 per 1M tokens); Meta's typical model costs less ($0.10 vs $3.44); OpenAI has the more accurate top model (95.8% vs 67.0% GPQA).
GPT and more
OpenAI
from $0.055 / 1M
39 models · typical $3.44 · top 95.8% GPQA, max reasoning
- Cheaper entry modelMeta
- More accurate top modelOpenAI
- More modelsOpenAI
The matchups that matter
| Meta | OpenAI | |
|---|---|---|
| Cheapest model | Llama 3.2 11B Vision Instruct $0.0175 via Lambda | gpt-oss-20b $0.055 via W&B Inference (CoreWeave) · 60.8% GPQA, medium reasoning |
| Most accurate (GPQA) | Llama 4 Maverick 17B 128E Instruct FP8 $0.0625 via Lambda · 67.0% GPQA | GPT-6 Astra $20.00 via OpenAI · 95.8% GPQA, max reasoning |
| Cheapest with 85%+ GPQA | no GPQA score yet | GPT-6 Luna $0.20 via OpenAI · 90.5% GPQA, max reasoning |
| Typical model price | $0.10 | $3.44 |
| Models we track | 23 | 39 |
Flagships: Llama 4 Maverick 17B 128E Instruct FP8 vs GPT-6 AstraCheapest: Llama 3.2 11B Vision Instruct vs gpt-oss-20bgpt-oss-120b vs Llama 4 Maverick 17B 128E Instruct FP8
Blended list price per 1M tokens (3 input : 1 output) at each model's cheapest provider; only models at least two providers sell. GPQA Diamond by Epoch AI (CC BY 4.0), best measured reasoning setting.
Every model, cheapest first
Meta
- Llama 3.2 11B Vision Instruct$0.0175 via Lambda
- Llama 3.2 3B Instruct$0.0175 via Lambda
- Llama 3.2 1B Instruct$0.02 via Novita AI
- Llama-3.1-8B-Instruct27.0% GPQA$0.0275 via Novita AI
- Llama Guard 3 8B$0.03 via Nebius Token Factory
- llama3-8b-instruct$0.0375 via DeepInfra
- Llama 4 Maverick 17B 128E Instruct FP867.0% GPQA$0.0625 via Lambda
- Llama 4 Scout 17B 16E Instruct51.8% GPQA$0.0625 via Lambda
- llama-v2-70b$0.10 via Fireworks AI
- llama-v2-7b$0.10 via Replicate
- llama-v2-7b-chat$0.10 via Replicate
- llama-v3-8b$0.10 via Replicate
OpenAI
- gpt-oss-20b60.8% GPQA, medium reasoning$0.055 via W&B Inference (CoreWeave)
- gpt-oss-120b75.8% GPQA, high reasoning$0.065 via W&B Inference (CoreWeave)
- GPT OSS Safeguard 20B$0.1025 via Amazon Bedrock
- GPT-5 Nano$0.1375 via OpenAI
- gpt-4.1-nano$0.175 via Replicate
- GPT-6 Luna90.5% GPQA, max reasoning$0.20 via OpenAI
- GPT OSS Safeguard 120B$0.2625 via Amazon Bedrock
- GPT-4o mini$0.2625 via OpenAI
- gpt-4o-mini-audio-preview$0.2625 via OpenAI
- GPT-5.6 Luna91.6% GPQA, max reasoning$0.45 via OpenAI
- GPT-5.4 nano$0.4625 via OpenAI
- GPT-5 Mini$0.6875 via OpenAI
Questions people ask
Is Meta cheaper than OpenAI?
Yes at the cheap end: Meta's cheapest model, Llama 3.2 11B Vision Instruct, starts at $0.0175 per 1M tokens, against $0.055 for OpenAI's gpt-oss-20b. The typical (median) model costs $0.10 at Meta and $3.44 at OpenAI.
Which has the more accurate model, Meta or OpenAI?
Meta's most accurate model we track, Llama 4 Maverick 17B 128E Instruct FP8, scores 67.0% GPQA (GPQA Diamond, Epoch AI) for $0.0625 per 1M tokens; OpenAI's GPT-6 Astra scores 95.8% GPQA, max reasoning for $20.00.
How many Meta and OpenAI models can I buy through APIs?
We track 23 Meta and 39 OpenAI models that at least two providers sell, with prices from each provider's pricing page.